[zeromq-dev] Disconnecting from permanently dead endpoints

[john skaller]

On 26/01/2012, at 12:32 PM, john skaller wrote:

> 
> Isn't it also a design fault? I mean, if you can connect/bind to multiple endpoints dynamically
> you obviously should be able to un-connect and un-bind.

API:

int disconnect(void *socket, void *connection);
int unbind(void *socket, void *connection);

Semantics:

When you connect or bind a socket, 0MQ associates the char* address
of the endpoint name with the internal representation of the endpoint.

When you call disconnect or unbind, ALL the endpoints which are
associated with that address are detached from the socket.

0MQ does not use the address for anything so it is perfectly safe
to free the buffer. It is also safe to use the same buffer for two connections,
in fact this is useful if you wish to disconnect or unbind them together.
if want to distinguish end points ensure the addresses passed in
are unique (i.e. if you're using wildcard connections make sure
to copy the buffer if you want to distinguish them).

No change to the existing C API is required for this extension.
No code is broken by this extension.

I guess the association of the address with the endpoint
internally is easy enough? The exact method of disconnecting
is open, but would be a bit similar to doing a close.  disconnecting
all sockets does NOT relieve the client of the need to close the socket.
I am guessing the disconnect would block in the same circumstances
as close would.

The return value is -1 on error, or the number of endpoints disconnected
or unbound. 0 is a valid result.  I do not know what errors might occur .. 
except perhaps trying to unbind a connection or disconnect a binding.


--
john skaller
skaller at users.sourceforge.net